Best practices when migrating from sklearn to torchnmf

[yoyolicoris]

fit

model = NMF(V.shape, rank=rank)
model.fit(V)

fit_transform

model = NMF(V.shape, rank=rank)
model.fit(V)
transformed_data = model.H.detach()      # the roles of H and W are swapped, which is different from sklearn

transform

model = NMF(V.shape, rank=rank)
model.fit(V)
model_new = NMF(W=model.W.detach(), H=(V_new.shape[0], model.H.shape[1]), trainable_W=False)
model_new.fit(V_new)

inverse_transform

V_hat = model(H)